Computing Mentorship in a So�ware Boomtown: Relationshipsto Adolescent Interest and Beliefs

�e Information SchoolUniversity of Washington

Sea�le, WAajko@uw.edu

Katie Davis�e Information School

University of WashingtonSea�le, WA

kdavis@uw.edu

ABSTRACTPrior work on adolescent interest development shows that men-torship can promote interest in a subject while reshaping beliefsabout the subject. To what extent do these same e�ects occur incomputing, where interest and beliefs have traditionally been neg-ative? We conducted two studies of the Puget Sound region in theUnited States, surveying and teaching 57 diverse adolescents withinterests in computing. In the �rst study, we found that interest incomputing was strongly related to having a mentoring relationshipand not to gender or socioeconomic status. Teens with mentors alsoengaged in signi�cantly more computing education and had morediverse beliefs about peers who engaged in computing education.�e second study reinforced this �nding, showing that teens whotook a class from an instructor who aimed to become students’teacher-mentor had signi�cantly greater positive changes in in-terest in computing than those who already had a mentor. �ese�ndings, while correlational, suggest that mentors can play a keyrole in promoting adolescent interest in computing.

KEYWORDSComputing education, mentorship, interest developmentACM Reference format:

DOI: 10.1145/3105726.3106177

1 INTRODUCTIONPrior work has shown that diverse adolescent populations aroundthe world continue to view computing as boring, antisocial, irrele-vant, male, and competitive [6, 8, 20, 24, 25, 35, 35–37, 44, 52]. Toaddress this, multiple e�orts worldwide aim to engage adolescentsin computing education with the goal of overcoming these views,developing interest in computing.

Sustaining this interest, however, requires more than just initialengagement. �eories of interest development, for example, viewinterest as something initially triggered by events such as an in-class

Permission to make digital or hard copies of all or part of this work for personal orclassroom use is granted without fee provided that copies are not made or distributedfor pro�t or commercial advantage and that copies bear this notice and the full citationon the �rst page. Copyrights for components of this work owned by others than theauthor(s) must be honored. Abstracting with credit is permi�ed. To copy otherwise, orrepublish, to post on servers or to redistribute to lists, requires prior speci�c permissionand/or a fee. Request permissions from permissions@acm.org.ICER ’17, Tacoma, WA, USA© 2017 Copyright held by the owner/author(s). Publication rights licensed to ACM.978-1-4503-4968-0/17/08. . .$15.00DOI: 10.1145/3105726.3106177

activity, and a�erwards, as something that is maintained over time,�rst externally by the learning environment and then individuallythrough intrinsic motivation, values, and identity [26]. From thisview, developing adolescent interest in computing is �rst a ma�erof triggering interest through experiences like Hour of Code or aparent enrolling a child in a computing camp and then a ma�er ofmaintaining and developing it through coursework, projects, andcommunity engagement.

Mentors can play a key role in both triggering and maintainingadolescent interest. �ey can devise learning experiences to sit-uationally maintain interest and they can also encourage furtherlearning, connecting learning to adolescents’ identities, shi�ingsituational interest into individual interest [26]. For example, stud-ies of technology �uency (which have investigated coding amongother technology skills) have found that parents can play mentor-ing roles in the development of technology skills, advancing theirchildren’s learning through parent-child collaborations [4]. �eserelationships are associated with adolescents’ deeper expertise andpositive a�itudes toward technology.

Parents, of course, are only one kind of mentor. Dawson, forexample, conceptualizes mentorship broadly [12], describing it asboth formal relationships between younger and older individuals,but also relationships between individuals of all ages with widelyvarying formality and levels of engagement. Lave and Wenger treatmentorship similarly, describing apprenticeship as a form of peermentorship that develops skill and knowledge [34].

Unfortunately, there is li�le prior work on informal mentoring[1]. Prior work has shown that many youth have some kind of in-formal mentor [5], and that these mentors can be central to youths’lives [32], but only a few works investigate computing. For example,Barron et al.’s investigation of parent-mentors only considered afew adolescents learning to code [4] and Ko’s study of computingautobiographies only reported a few students mentioning mentor-ing relationships as part of their developed interest in computing[33]. Most prior work has instead focused on formal mentoringin the workplace (e.g., [28, 29]) and in educational contexts withat-risk learners (e.g., [38, 51]. �is trend is also true of researchon mentoring in computer science in higher education, which hasfound that strong communication skills and an appropriate person-ality �t are key [39, 42, 49], mirroring more general research onformal mentoring [23, 30]. By following these practices, CS facultycan in�uence enrollment decisions [45], increase retention [10, 21],and produce more e�ective learning [16]. Formal peer mentoring,in contrast is fraught with challenges [27], but can improve learners’sense of community [14].

Amy J. Ko

Amy J. Ko and Katie Davis. 2017. Computing Mentorship in a So�ware Boomtown: Relationships to Adolescent Interest and Beliefs. In Proceedings of ICER ’17, Tacoma, WA, USA, August 18-20, 2017, 9 pages.

ICER ’17, August 18-20, 2017, Tacoma, WA, USA

While all of this prior work suggests that mentorship may be ane�ective contributor to developing adolescents’ interest in comput-ing, there has been li�le systematic investigation into the relation-ships between computing mentoring, interest, and beliefs amongyouth. Speci�cally:

• RQ1: To what extent is mentorship related to positivebeliefs about computing, interest in computing, and en-gagement in computing education?

• RQ2: What relationship does acquiring a computing men-tor have with interest in computing?

Studying these question is complex. As Renninger and Hidi dis-cuss in their review of conceptualizations of interest, researchershave only just begun to develop systematic, consistent conceptual-izations of interest [43], providing few methods or measurementsthat control for interacting roles of age, identity, cognitive devel-opment, and learning context. Moreover, conducting longitudinalcontrolled experiments on the e�ects of mentorship is not onlychallenging, but premature, without �rst establishing the potentiale�ects of computing mentorship. �erefore, our goal was to mea-sure these potential e�ects and do so in a community near universalaccess to computing education and an abundance of potential com-puting mentors, partially mitigating confounding factors such asaccess to computing education and availability of mentors. Whilethese methods cannot show that mentorship was a cause of thebeliefs and interests we observed in either study, we hope thatthey provide insight into the potential e�ects of mentorship incomputing education, laying a foundation for further investigation.

2 APPROACHOur approach to investigating these two questions was to studyadolescent experiences with computing mentorship in the Sea�lemetropolitan area in Washington state in the United States (alsoknown as the Puget Sound region). Sea�le is unique in that whileit only has tens of thousands of students, it also has over 90,000professional so�ware engineers, possibly providing most studentswith access to potential computing mentors. Moreover, public andprivate schools in the region also provide near universal access toboth formal computing education in secondary school, each o�er-ing at least one if not more computer science courses, some withdedicated CS instructors, some who primarily teach math, science,or technology, and some who are industry volunteers through pro-grams like TEALS (Microso�’s Technology Education and Literacyin Schools program). �e city also is full of informal learning op-portunities, including adolescent coding camps designed to developinterest. Dozens of companies including Microso�, Google, andZillow also sponsor frequent day or week-long computing educa-tion events for teens across the region. Finally, Sea�le is unique inthat it is also one of the most highly educated and wealthy citiesin the U.S. and yet still contains socioeconomic diversity due tothe signi�cant in�ux of immigrants and refugees relative to its size.�ese conditions made it possible to explore more idealized con-ditions for mentorship, mitigating structural inequities in accessto learning and mentorship, while also preserving socioeconomicdiversity and a degree of racial diversity and segregation. �erefore,our exploratory study is as much an investigation into the uniqueconditions in Sea�le as it is a study of mentorship.

In this context, we conducted two studies. First, we surveyedtwo socioeconomically and racially diverse groups of high schoolstudents. One group of students enrolled in a 1-week, half-dayweb design course through a fee-based summer camp programsponsored by the University of Washington, which tends to a�ractupper-middle class White and Asian students from Sea�le’s wealth-ier neighborhoods and suburbs. �e other group was enrolled ina federally-funded Upward Bound program, a �rst-generation col-lege preparation program that recruits from three of south Sea�le’spublic high schools, which tend to enroll racially diverse immigrantyouth living in neighborhoods with low socioeconomic status. Weasked students about beliefs, interests, and mentorship relation-ships. We then taught the Upward Bound students in a six-weekcourse that framed the instructor as an informal computing men-tor, investigating the potential for his mentorship to strengthenstudents’ situational interest in computing.

3 STUDY 1: MENTORSHIP, BELIEFS,ENGAGEMENT, AND INTEREST

�e goal of our �rst study was to investigate adolescents’ computingmentoring relationships, their beliefs about computing, and theirinterest in computing.

3.1 Method3.1.1 Population and Sampling. Our target population was ado-

lescents aged 14-18 living in Sea�le or the broader Puget Soundregion. Our objective was to recruit as diverse a group of teens aspossible, across race, age, gender, socioeconomic status, geogra-phy, and school. To do so, we o�ered two types of summer codingclasses. One was a university-based Upward Bound program. Up-ward Bound (UB) is a federally funded college preparation programthat helps high school students who are low-income and/or have noparent with a bachelor’s degree enter college. �ere are currently826 programs in the U.S., many of which have existed since the 1964Economic Opportunity Act that founded them. �e program weworked with serves three urban south Sea�le public high schools.�e program serves about 125 students per year. �e program isfree; students receive lunch money and a stipend to a�end. In 2016,79% were both low-income and �rst-generation immigrants, 50%identi�ed as female, 35% as South Asian, 19% as African, 16% asAsian, 14% as Hispanic/Latino, 10% as Black, 4% as two or moreraces, and 2% as White. �e program’s high school graduationrate is 98% and its college graduation rate is consistently above60%, with many alumni pursuing graduate studies. �erefore, theprogram primarily serves the high-achieving end of Sea�le’s lowerincome schools.

�e UB program o�ers a full summer curriculum that includesa�ernoon electives. We o�ered a course titled “Web Design” withan enrollment limit of 25. �e course description made no men-tion of “coding.” Students could choose between it and electives inballroom dancing, swimming, or music. Program administratorssolicited student preferences and then randomly assigned studentsto their 1st and 2nd choices, ensuring a balance along racial andgender lines; however, administrators also encouraged students tofocus on classes that would satisfy graduation requirements, andso most students who enrolled in the class had an existing interest

Amy J. Ko and Katie Davis

Computing Mentorship ICER ’17, August 18-20, 2017, Tacoma, WA, USA

Neighborhood

AsianWhite

White/AsianBlack

Hispanic Hispanic/Asian

White/Pac. IslanderArab

EnglishSpanishChinese

VietnameseKoreanFrench

Also Japanese, Arabic, Thai, Laos, Amharic, Cham, Hindi,

Tamil, Italian, Malayalam

doctoratemasters

some graduatecollege

some college high

some high middle

some middle some elementary

software developerprofessor

retireeengineer

doctorhomemaker

Boeing employee teacherdentist

nurseunemployed

Also unknown (2), social worker (2), custodian (2), housekeeper (3), business owner (2), lawyer (2),

accountant (2) telecommunications (2), business person (2), health care manager (2), veterinarian, statistician,

tax preparer, academic counselor, financial aid officer, chemist, biologist, contractor, interior designer, speech

therapist, flight attendant, database administrator, sales associate, manager, grocery worker, researcher, data

analyst, microbiologist, fashion designer, Amazon employee, librarian, student, chef, manicurist,

construction worker, landscaper, hairdresser, house flipper, waitress, and seamstress

RAC

IAL

IDEN

TITY

LAN

GU

ANG

ESPA

REN

TS’ E

DU

CAT

ION

PARE

NTS

’ OC

CU

PATI

ON

S

NEI

GH

BORH

OO

D Also Tumwater RentonMukilteoKenmore

Figure 1: Student demographics. Students could choosemultiple races and languages, one parent education level, one city, andup to two guardian occupations. Dark circles and bold text are SY students, white circles and non-bold text are UB students.

in computing. �e UB class was 6 weeks, 4 days per week, and 36hours total. We describe the class in detail later.

Our 2nd class was two sections of a 5-day, 3-hours per day codingcamp (also titled “Web Design” and using the same UB coursedescription) o�ered through a university-based summer youth (SY)program. �is program markets to teachers in Sea�le’s wealthiernorth suburbs and the eastern suburbs of Bellevue, Redmond, andKirkland. �ese classes tend to a�ract youth from upper-middleclass families. Teachers distribute camp information to parents,who enroll their teens. �e course had a registration fee of $275for the week, providing a barrier to enrollment for lower-incomestudents (though the program had a limited budget for �nancialassistance). Our class’s fee was comparable to other half-day campsin the city. We o�ered two sections: one in the morning and one inthe a�ernoon, with an enrollment limit of 25 each.

3.1.2 Participants. We successfully enrolled 57 teens across bothclasses. We �lled both SY camps, but had several last minute can-cellations, resulting in 44 total enrolled students. Our UB class hadroom for 25, but we only enrolled 13, two of whom dropped theclass in favor of ballroom dance, but still �lled out our survey.

Despite the enrollment challenges, our sample was still diversealong many dimensions. Students were 39% female and aged 14-17 (median 15). As shown in Figure 1, students were all �uentin English, but most were bilingual (12 reported never speakingEnglish at home), most identi�ed as White or Asian, many hadhighly educated parents but across a diversity of occupations, andstudents came from across the region, but primarily Sea�le.

As with any sample, there were several biases. All students re-ported wanting to go to college, despite Sea�le Public’s 2015 highschool dropout rate of 22%. Over 75% of students had a parentwith a bachelor’s degree, which is higher than the city’s 2015 countof 54% of residents over age 25 with a bachelor’s degree. (Sea�lehas one of the highest education rates of any U.S. city, and so theregion already skewed toward high educational a�ainment.) Seat-tle’s racial demographics are also unique, with many low-incomerecent Asian immigrants, and many a�uent Hispanic families. �eresulting sample is therefore re�ective of the city’s mix of educatedprofessional families and its recent immigrant families.

3.1.3 Data Collection. We began all classes by having students�ll out a web-based survey. Both classes were held in adjacentcomputer labs on a university campus, roughly identical in layoutand identical in hardware. Students took approximately 30 minutesto complete the survey. It began with demographic information (agein years, neighborhood they lived in, school and grade they wouldenter in the fall). Next, to obtain data on students’ interests broadly,we asked students about their academic plans, asking whether theyplanned to go to college and measuring students’ possible selves[41], responding to the prompt “Describe your vision of your life atthe age of 25, assuming everything goes well. Who are you? What willyou have achieved? What will your goals be?” �is question allowedus to understand students’ interests, identities, and motivations.�e survey then probed for interest in computing (detailed shortly),followed by beliefs about peers engaged in computing educationand about so�ware developers.

ICER ’17, August 18-20, 2017, Tacoma, WA, USA

�e end of the survey probed identity (to avoid potential stereo-type threats [47], as questions about identity can in�uence howrespondents describe their interests, prior knowledge, and abil-ity). We asked for students’ languages that they speak �uently,ordered from most to least �uent. We also asked them to notewhich language they spoke most at home. We then asked for racialidentity (following the latest U.S. census recommendations foundat census.gov/topics/population/race/about.html), gender, and socioe-conomic status. Measuring socioeconomic status (SES), which isviewed as a combination of education, income, and occupation,is non-trivial [19]. However, there is evidence that parents’ ed-ucational a�ainment is highly predictive of youth SES in adult-hood [17]. �erefore, following the latest best practices on SESmeasurement [31], we asked students “How many years of schoolhas your most educated parent or guardian completed (0-25 years)?(0=no school, 5=�nished elementary school, 8=�nished middle school,12=�nished high school, 16=�nished college, more than 16 meansgraduate school).” We also asked students to identify their parentand/or guardians’ occupations to help us understand their parents’occupational proximity to the so�ware industry.

3.2 Results3.2.1 Access to ComputingMentors. We �rst consider the degree

to which students reported access to potential computing mentors.Of all students, 49 (89%) reported knowing someone who knew”how to code.” �is did not vary by SES (F (1,55)=.86, p=.36), gender(Fisher’s, p=.72), or UB/SY class (Fisher’s, p=.37). Students reportedtheir relationships with these individuals as friends (17), fathers(16), teachers (9), brothers (4), uncles (3), cousins (2), a mentor, aneighbor, a friend’s sibling, a parent’s friend, a sister, and a tutor.

To analyze the nature of students’ relationships with these in-dividuals, we analyzed the students’ descriptions for qualities ofmentoring. Although the literature on mentorship has not yetagreed upon a theoretical account of mentoring, there are fourfacets that consistently arise in the mentoring literature [11]: 1)psychological and emotional support, 2) support for goal se�ingand career selection, 3) teaching of academic subject knowledge,and 4) framing of the mentor as a role model. We liberally opera-tionalized computing mentoring relationships as any relationshipthat a student described that included one or more of these facets,including descriptions of explicit encouragement, learning activi-ties, including explicit instruction, help enrolling in classes, visits toworkplaces, role modeling and the fostering of other mentoring re-lationships. We treated all other relationships (including those withindividuals who knew how to code) as non-mentoring relationships(for example, many described fathers, friends, and brothers whonever explained their so�ware development jobs or a�empted toteach them anything about computing). Based on this de�nition,24 (42%) of the students described having at least one computingmentoring relationship. Whether a student had a mentor did notdi�er by SES (F (1,55)=.87, p=.36), gender (Fisher’s, p=.17), or UB/SYclass (Fisher’s, p=.2).

Students with mentors described a wide range of mentoringrelationships:

“My dad is a professor in biology here at the university and he hasbeen teaching me python recently.” (SY, White male, 16)

“One of my best friends is very interested in programming, and hastaught me some basics about HTML5” (SY, White male, 14)

“My dads friend sister works in Microso� and he’s a developer weonly got to talk about her job a li�le bit, she took me on a tour ofthe campus and saw what people do there. it was pre�y cool” (UB,Black female, 17)

“In the past few summers, I had a private computer programmingtherapist… She taughtme new programs and new coding. She didn’tdo it this year because now she’s going to college.” (SY, White/Asianmale, 14)

“My mom is a website developer, she teaches me about the tools sheuses. My dad, writes code in java. My friends taught me how toprogram also” (SY, White male, 17)

“My cousin wants to be a computer science major in college andruns a club at her school called “Girls Who Code.” She encouragesme to try coding. ” (SY, White/Asian male, 16)

“My dad is a so�ware engineer and he frequently talks to me abouthis job. He has enrolled me in several classes and in our free time,he o�en teaches me. ” (SY, Asian female, 14)

“My neighbor Laura who did APCS and really enjoyed it and intro-duced me to it.” (SY, Hispanic female, 15)

3.2.2 Beliefs about Computing Education and So�ware Develop-ers. Next, we turn to the beliefs that students reported about com-puting and computing education. We �rst asked students, Whatkinds of students take your school’s technology courses? In theirresponses we saw few of the beliefs reported in prior work (e.g.,[3, 36]), in which adolescents a�ributed student interest in com-puting to gender or intelligence. Organizing their free responsesby the speci�c words they used, we found that students describedstudents who engaged in computing education as “anyone” (17),“people who love computers” (9), “overachievers” (6), “people whowant good jobs” (5), “people who are required to” (5), “slackers” (3),“Asians” (2), “boys,” “girls,” “cool people,” “game lovers,” and “loners.”Eight students said they were unsure about who takes CS becausethey were new to their school.

When we considered the beliefs reported by students with andwithout mentors, there were only minor di�erences. Of studentswithout mentors, 33% described students who took computingcourses as “anyone” and 18% were unsure, whereas of the studentswith mentors, only 25% reported “anyone” and 8% were unsure.However, the students with mentors had more diverse beliefs thatdi�ered from those without mentors, including “boys,” “girls,” and“cool people.”

In addition to beliefs about their peers who engaged in comput-ing education, we also considered students’ beliefs about profes-sional so�ware developers, asking students “What characteristicsdo you think someone must have to be a so�ware developer?” Adjec-tives used by more than 10% of students without mentors included“creative,” “patient,” “smart,” “hard-working,” “intelligent”, and “per-severant,” all mentioned by more than 10% of students. Studentswith mentors used largely the same language, but unlike studentswithout mentors, also used the adjectives “logical”, “collaborative,”“enthusiastic,” and “precise.”

Amy J. Ko and Katie Davis

Computing Mentorship ICER ’17, August 18-20, 2017, Tacoma, WA, USA

We also asked students “What do you think so�ware developersdo at work?” Most students, regardless of whether they had a com-puting mentor, believed that developers “code all day,” “go to a fewmeetings,” and “think of product ideas.” Most students did not de-scribe the collaborative aspects of the job, portraying a typical dayas a solitary one. We also asked students “What must someone doto become a so�ware developer?” Responses universally mentionedtaking courses and practicing extensively, as in this representativeresponse: “�ey need to learn about coding and practice it until itbecomes as easy as breathing” (UB, Vietnamese male, 15).

3.2.3 Engagement in Computing Education. Next we considerstudents’ engagement in computing education and its relationshipto students’ access to computing mentors. We asked students to listall of the technology courses that their school o�ers, what kindsof students take those technology courses, and whether they hadtaken those courses. We also asked if they had used other onlinelearning technologies to learn to code.

Of the 57 students, 50 (88%) could name at least one computingcourse in their school. �ere were no signi�cant relationshipsbetween the number of courses known and SES (r(55)=.07, p=.60),gender (F (1,55)=2.5, p=.12), UB/SY class (F (1,55)=.84, p=.36), orwhether a student had a computing mentor (F (1,55)=0.24, p=.63).We con�rmed that the schools of the 7 students who could not namea computing class did o�er one in the year prior; they were just notaware of it because they were incoming freshmen or had recentlymoved. Of these 7 students, 6 did not report having a computingmentor.

We asked students what CS courses or learning experiencesthey had engaged in. Twenty-seven (47%) reported having alreadyengaged in some kind of computing education. Of these students,89% (24 students) had taken an elective class at school. Who hadtaken a course did not di�er by SES (F (1,55)=1.1, p=.29), gender(Fisher’s, p=.10), or UB/SY class (Fisher’s, p=1.0). Courses taken diddi�er signi�cantly by whether a student had a computing mentor(F (1,55)=4.44, p=0.04); 16 students (67%) with a mentor had taken atleast one class (and only 3 of these reported that their sole mentorwas their teacher), compared to 27% of students without a mentor.

Only 4 students mentioned engaging in informal learning. Oneread a Python book over the summer; one started but did not com-plete a Codecademy Java tutorial; one started but did not completea Codecademy JavaScript tutorial; and one had completed a seriesof Udacity online classes and Codecademy tutorials. �ree of thesefour students reported having a computing mentor.

Despite only half of the students mentioning that they had en-gaged in computing education, when we asked students to listthe programming languages that they had encountered, 42 (74%)mentioned having encountered at least one. Students mentionedfamiliarity with Scratch (28), HTML (19), Java (18), JavaScript(12), Python (12), CSS (8), Excel (8), C/C++ (6), Minecra� mods(7), Kodu (4), C# (2), Processing, Alice, Ruby, Go, Swi�, Lua, andPHP. �ere were no trends in who had encountered a languageby SES (F (1,55)=1.2, p=.28), gender (Fisher’s, p=.22), or UB/SY class(Fisher’s, p=.29). However, students with computing mentors re-ported encountering signi�cantly more programming languages

(F (1,55)=11.4, p=0.001), with 92% of students with a mentor encoun-tering at least one language, compared to only 40% of studentswithout a mentor.

3.2.4 Interest. We now turn to students’ interest in comput-ing. According to our adopted theoretical framework [26], stu-dents’ engagement in computing education and access to mentorswould have triggered and maintained interest in computing. Wewould therefore expect that there would be a strong relationshipbetween mentorship, engagement in computing education, andinterest. Prior work would also suggest a relationship betweengender and cultural factors, mediated by beliefs and access [36, 37].

To investigate interest, we �rst analyzed the possible selves [41]that students described, inspecting them for interests. Although13 students wrote that they were unsure about their careers and15 mentioned computing careers (so�ware developer (8), gamedeveloper (5), and web developer (2)), the other 30 described diverseinterests: including doctor (6), writer (2), mechanical engineer (2),aeronautical engineer, autism activist, bioengineer, businessman,computer engineer, drug researcher, Ethiopian politician, illustrator,and several other distinct professions. �is variety suggests thatmost of the students did not enroll in our courses because theywere explicitly interested in computing as a career.

To measure students’ interest in computing, we adapted the scaleused by Oh et al. [40], presenting the following �ve items on a7-point Likert scale: 1) I’m interested in taking courses that help melearn to code, 2) I am interested in careers that allow me to use codingskills, 3) I would like to learn to code because it will help me preparefor college, 4) I would like to learn to code because it will help me geta good job, and 5) I would like to learn to code because it will help mecreate new technologies. We mapped each response to a -3 to 3 scaleand computed the mean of the �ve items.

Interest in computing skewed positive. �e mean response was1.1 and ranged from -2 to 3, suggesting most teens viewed comput-ing as an interesting, valuable subject to learn. Only 10 students(18%) had a�itudes below 0 on the scale, including 5 of 11 (45%)of the UB students and 5 of 46 (12%) of the SY students. �is wasconsistent with the reasons that students listed for enrolling, whichincluded: “I’m interested” (28), “parents wanted me to” (9), “seemeduseful” (7), “curious about the topic” (3), “friends encouraged me”(2), “bored this summer” (2), “placed by counselor” (2), “retake” (2),“avoiding physical activity,” and “for credit.”

To test the relationship between mentorship and interest in com-puting, we built a linear regression. For predictors, we includedgender, since prior work has observed signi�cant disinterest fromgirls because of the culture of computing education (e.g., [37]). Weincluded number of programming languages encountered as anindicator of prior learning about computing. We included SES, asprior work has shown that higher SES is related to higher aca-demic ambitions [46]. And �nally, we included whether the studentreported a mentoring relationship, given its clear relationship tointerest reported in prior work [4, 26] and many students had de-scribed it as a signi�cant factor in their interest. �e interest scaleand the predictor variables satis�ed the assumptions for a linearregression.

Table 1 shows our resulting model, which explained a signi�cantproportion of the variance in interest (R2=.223, F (4,52)=3.72, p=.01).

ICER ’17, August 18-20, 2017, Tacoma, WA, USA

Table 1: Linear regression predicting interest. *=p<.05.

B SE B β

Gender .234 .316 .098SES .097 .070 .177# PL encountered .131 .071 .254Had mentor .623 .319 .265*

Gender, SES, and programming languages encountered did notexplain a signi�cant proportion of variance in interest, but havinga computing mentor did, and was responsible for a 0.623 increaseon the -3 to 3 computing interest scale.

4 STUDY 2: FORMING COMPUTINGMENTORING RELATIONSHIPS

�e results of Study 1 suggested that the most signi�cant factorrelated to interest in computing was mentoring. However, thisstudy did not allow us to observe mentoring relationships beingformed, or the possible e�ects of acquiring a mentor on changein interest. �erefore, in Study 2 we taught the 6-week UB classwith the explicit goal of the instructor (the 1st author) to developinformal computing mentoring relationships with each of the 11students. To do this, the instructor aimed to give the students anaccurate portrayal of the web development communities of practice[34], while developing a sincere interest in each individual student’slearning, leveraging his decade of experience as a teacher and 5years of experience as a professional web developer.

4.1 Course Design�roughout the course, we a�empted to follow best practices fromeducation and computing education. �e class was only 11 stu-dents, reducing the negative e�ects of large class sizes (e.g., [2]).�e instructor followed the best practices of classroom manage-ment [18]. For example, because the class was in a computer lab,he managed student a�ention by creating separate spaces for lec-ture and computer-based learning to prevent students from beingdistracted. He established and enforced clear rules of conduct, pre-venting disruptions. He also followed the evidence-based practiceof learner-centered teacher-student relationships [9], a�emptingto create authentic relationships in which students were trusted,given responsibility, spoken to honestly and warmly and treatedwith dignity. �is approach was consistent with the tone of in-struction throughout the rest of the UB program. In addition, theclass followed NCWIT’s inclusive learning practices (ncwit.org/resources/type/promising-practices), explicitly discussing stereotypethreat, imposter syndrome, and theories of intelligence in the con-text of the so�ware industry, computing education, and in theclassroom. �e instructor also followed NCWIT’s practice of inten-tional role modeling (based on studies such as [48]), 1) explainingwhat made his role as a practitioner relevant to their learning, 2)describing his personal history and how it related to the students’experiences, 3) speaking about his strengths and weaknesses andhow they related to his expertise, and 4) showing them how hea�ained his position.

Our instruction was also informed by communities of practicetheory [50], as prior work has explored in computing education[13, 22]. In this theory, a newcomer’s purpose is to learn to talk anddo as the community does, acquiring the norms, practices, skills,and tools that the community uses to do its work. In our class, wefollowed the practices described in prior work (e.g., [7, 22]), havingstudents use authentic web development tools of GitHub, JSFiddle,and Bootstrap, but sca�olding them to facilitate learning. We intro-duced the students to three experienced web developers who hadworked at local companies. We showed the students the diversityof so�ware companies in the Puget Sound region, teaching themabout the dozens of large companies and their di�erent corporatecultures and values, but also the hundreds of so�ware startups.Across four reading assignments, students re�ected on equity incomputing education, reading the White House press release onthe CS for All initiative, Chapter 4 of Stuck in the Shallow End [36](which covers preparatory privilege), a viral blog post by a femaleso�ware engineer on imposter syndrome in the so�ware industry,and a case study wri�en by the 1st author about his experiences ata so�ware startup.

We designed a 3-week project in which the students designed anddeveloped a website individually or in teams, creating somethingpersonally meaningful. �e class involved 6 projects across the 11students. �ey chose diverse topics, including an informational siteabout Ethiopian culture, practical applications of the philosophy ofethics in everyday high school life, a youth book recommendationsite, a site to help high school students reduce stress and avoidprocrastination, a site for sharing trends in athletic shoes, a sitefor sharing life hacks that make high school easier, and a site forsurviving junior-year humanities courses. �roughout 3 weeks ofweb development, the instructor provided individual help, o�eringconstructive, guiding feedback about each student’s e�orts. He alsorequired students to provide weekly peer evaluation about theirteammates’ collaboration skills.

�e instructor taught students about pathways to joining webdevelopment communities of practice, giving detailed informationabout what kind of education was required, what types of jobs andcompanies hire web developers, and what kinds of people currentlypursue these pathways, including his own path. �e web developerswho visited the class also talked about the pathways that they took,what they regre�ed about their path, and what surprised them asthey followed their path.

In the context of the pedagogical strategies and content de-scribed above, the instructors’ mentorship formation strategiesincluded the following: 1) talking each day to each student abouttheir progress on learning, 2) in these conversations, explicitly link-ing their progress to the goals they reported in their possible selves,and 3) at the end of the course, having an explicit mentoring con-versation with each student individually, o�ering to help them withtheir college applications, connect them with further computingeducation experiences, and answer their questions over email asthey approached college.

Finally, at the end of the course, the instructor gave students thesurvey from Study 1, excluding redundant demographic questions.

Amy J. Ko and Katie Davis

Computing Mentorship ICER ’17, August 18-20, 2017, Tacoma, WA, USA

Table 2: An ordinal regression predicting change in interestin computing. *=p<.05.

B SE B Wald β

Interest before class -0.46 0.25 3.49 .06Male 4.05 2.27 3.18 .07SES -0.21 0.58 0.13 .71# PL encountered -0.67 0.67 0.99 .32No mentor 5.79 2.82 4.21 .04*Midterm -0.43 0.19 4.97 .16

4.2 ResultsUB student interest in computing changed signi�cantly, movingfrom a mean of 0.5 to 1.35 (t(10)=2.9, p=.016). None of the studentsbegan at the top of the interest scale and all but one had an increasetoward neutral or positive interest.

Students’ descriptions of their possible selves at the end of theclass revealed some reasons for these changes. Six of the 11 studentswere still sure they wanted to be doctors, lawyers, so�ware devel-opers, and pharmacists. However, 5 of the 11 students describedsubstantially di�erent possible selves, incorporating computinginto their identities. One student who was particularly engaged inclass went from being generically interested in medicine to want-ing to strengthen her coding and drawing skills, both of which sheexplored in the design and implementation of her team’s website.One went from wanting to be a doctor to wanting to �rst studycomputer science to make money, then use the money to go tomedical school to study ophthalmology. Another student who wasparticularly adept at learning JavaScript went from describing hisfuture as “a blank canvas” to wanting to study computer scienceand aviation, because “perhaps those �elds need someone to writeprograms to analyze data and make be�er equipment” (UB, Asianmale, age 15). Yet another student came into the class with strongweb development skills from a prior course but low self-e�cacydue to a bad teamwork experience. A�er working independentlyon his project, he le� the class con�dent in his skills and said hewas now considering computer science as a major.

To investigate what was related to these changes, we built aregression similar to that in Study 1, but this time predicting changein interest from several factors. We included the same factors frombefore: number of programming languages encountered before theclass as an indicator of prior knowledge, gender, SES, and whethera student had a mentor. However, we also included interest incomputing prior to the class (as this would likely contribute tointerest a�er the class) and the students’ scores on their �nal exam(as success in CS classes can greatly in�uence interest [33]). Changein interest violated normality assumptions for a linear regression,and so we built an ordinal regression model instead, computing thedi�erence in the pre and post sum of Likert items.

Table 2 shows the resulting model. �e model explained a sig-ni�cant amount of the variance in change in post-class computinginterest (χ2=21.8, p=.001, Cox and Snell R2=.86). In this model,not having a mentor before the class was signi�cantly related topositive changes in interest in computing.

When we asked the students to explain how the class in�uencedtheir interest in computing, if at all, their explanations includedmany references to the instructor, and particularly to the inclusivelearning environment:

“�is class helped me understand what jobs involved with techwould look like and how it’s not just a lonely person in the base-ment.” (UB, Asian male, 15)

“�e readings from this summer encouraged me to acquire technol-ogy skills since President Obama plans to enforce Computer Scienceclasses and it seemed to be possible for anyone to learn once theyget past imposter syndrome like that one woman in the reading.”(UB, Asian male, 16)

“�is class has greatly in�uenced my a�itude towards coding. Itallowed me to look past the stereotype that I never realized.” (UB,Asian male, 16)

”You de�nitely changed what I thought about computer science andweb design. I love designing things, but I never thought one day Icould design my own website in 6 weeks. �e environment and classmade class more enjoyable for me. �is environment made me wantto be part of it. Unlike classes where you feel like you aren’t needed.Lastly, you taught me the thought and idea of what technologyskills can do for you in the future even if it’s in the medical �eld.�at makes me want to continue learning.” (UB, Asian female, 16)

Students’ re�ections on the reading assignments revealed similarshi�s in interest. For example, one of the readings was a blog postby a woman who started as a web developer and faced impostersyndrome, but eventually realized that failure was part of everydeveloper’s job. Two girls were surprised:

”Learning to code is intimidating because it’s like learning a foreignlanguage…When I feel frustrated, I start to feel like I don’t belong inthe class. I start to question whether I should pursue a career whereI would never be able to master a subject. Technology is alwayschanging, and I have to ask myself if I would feel comfortablelearning something new every day for the rest of my life. I nowknow it’s normal to feel uncomfortable when you are learningsomething new and that is why I believe I will have the perseveranceto learn as much as I can about web design.” (UB, Asian female, 15)

”When I �rst took web design, I thought the guys would be expertsin this. Because I know guys who knew a lot about computer [sic]and how it works. So in the beginning I was a bit intimidated untilI learned that all of us know nothing about web design. It made meless scared when I want to ask a question about web design.” (UB,Asian female, 16)

�ree of the students that did not have mentors prior continuedto engage the instructor as a mentor a�er the class ended. Onesought advice on computing-related summer jobs and at the timeof this writing is participating in a Girls Who Code camp as aninstructor upon the instructor’s recommendation. Another soughtadvice on how to engage in computing research as an incominguniversity freshman and is now participating in a �rst-generationcollege student research mentorship program. A third sought adviceon what to study to combine computing and medicine.

ICER ’17, August 18-20, 2017, Tacoma, WA, USA

5 DISCUSSIONOur results contribute the following discoveries about Puget Soundteens with interests in computing:

• Teens can have mentoring relationships with a range ofpeople, including friends, parents, siblings, cousins, teach-ers, and even neighbors.

• Teens’ beliefs about peers who engage in computing ed-ucation were mostly positive, describing high-achievingyouth of any gender, race, or ethnicity.

• Teens’ beliefs about so�ware developers described devel-opers as creative, patient, intelligent, hard-working, andperseverant people.

• Beliefs held by those with mentors were more diverse andhad greater depth than those without mentors.

• While most teens knew about computing education oppor-tunities regardless of whether they had a mentor, havinga mentor was related to engaging in more of them andencountering more programming languages.

• While most teens expressed interest in computing as a skill,subject, and career, teens with mentors reported strongerinterest regardless of gender or socioeconomic status.

• Teens that reported the �rst author as a teacher-mentorreported signi�cantly higher changes in interest in comput-ing than students that already had mentors, regardless ofgender, socioeconomic status, performance in class, priorknowledge, or prior interest.

�ere are many ways to interpret these results. First and fore-most, none of these results are causal. It is possible, for example,that students who developed interests in computing did so withoutmentors (for example, through Code.org’s Hour of Code or someother compulsory learning se�ing) and then sought mentors in theirsocial networks. It is also possible that interest was entirely men-torship driven, with mentors proactively triggering the students’interest in computing and then maintaining it through subsequentactivities. Students’ self-reports about their mentoring relation-ships suggest the la�er was more likely than the former, as moststudents with mentors described fathers and siblings proactivelyintroducing them to computing.

Similarly, our e�orts to form a mentoring relationships with theUB students may have not been the cause of students’ increasedinterest in computing. Students may have had other experiencesin the summer that developed interest in computing, such as thecollege prep class that encouraged the UB students to develop theircareer interests. It may have been the combined e�ect of mentorshipand career interest re�ection that produced the increases in interest.�e ability to form a mentoring relationship with the Asian studentsmay have also been mediated by the somewhat Asian appearanceof the instructor. �at said, prior work on role models suggests thatidentity ma�ers less in recruiting than in retention [15].

Our study has limited generalizability, �rst and foremost becauseof the sampling biases in our data. At the time of our study, therewere tens of thousands of high school students in the Puget Soundmetropolitan area and we only learned about 57 of them. Ourclaims are limited to the types of students who enroll in UB (lowincome, �rst-generation college students), and the types of studentswho enroll in the SY camps that we o�ered (students with highly

educated parents interested in their teens learning to code). We didnot study any low-income students who were not college bound,nor did we study middle-class students who did not have su�cientprior interests (or parents with su�cient interest) to enroll in ourweb design courses. �ere were also likely structural inequitiesthat prevented the full population of students of being aware ofour classes, limiting participation from teens in lower SES, lowereducational a�ainment families.

�ere is also li�le evidence to claim generality of our results toother regions in the world. For example, even regions that sharesome features to the Puget Sound such as Silicon Valley have otherfeatures that could have changed our �ndings: it is more diverseracially and socioeconomically than Sea�le, not to mention anorder of magnitude larger in population. Moreover, it is unclearhow our results would generalize to regions that do not have vibrantso�ware industries or universal access to computing education.

Our measurements relied on self-report surveys and scales thatcan be sensitive to the timing of self-report, introducing somethreats to construct validity. Moreover, while most of the measure-ments in the surveys were taken from validated scales, not all ofthem were, and so some of the data may not accurately re�ect thephenomena we intended to study. Moreover, we used regressionsto investigate predictors of interest, which only suggest correlation.

In light of these multiple interpretations and limitations, the needfor future work is extensive. Studies should investigate the causale�ects of mentorship, for example, through formal mentorshipprograms, exploring the potential for virtuous cycles of mentorship-triggered interest and learner-driven interest maintenance. Futurestudies should investigate the granular e�ects of speci�c forms ofmentorship, such as the explicit instruction by parents and siblingsthat students reported. Studies should also investigate the speci�ce�ects of mentorship on other constructs such as changes in identityand shi�ing of beliefs, especially in se�ings with the more negativebeliefs we did not observe.

In addition to developing a be�er theoretical account of comput-ing mentorship, there are also numerous design questions abouthow to structure mentorship in computing. For instance, are adults,siblings, friends, or teachers be�er or worse positioned to men-tor adolescents about computing? Are teacher-mentors scalable,given typical class sizes in public schools? What kinds of skills domentors need to be credible and relevant to teens? Would recenthigh school graduates pursuing computing be e�ective mentors? Isremote mentorship feasible, especially in light of the majority of re-gions in the world having substantially fewer mentors than regionslike the Puget Sound? �ese alternatives suggest that increasinginterest in computing, even in the presence of universal access tolearning opportunities, may require signi�cant investments fromthe computing community to increase interest and engagementin computing education. Recent e�orts like NextBillion.org, whichhelps connects people with disabilities to industry professionals isone example of what such e�orts might entail.

ACKNOWLEDGMENTS�is material is based upon work supported by Microso�, Google,Adobe, and the National Science Foundation under Grant No. 1539179,1314399, 1240786, and 1153625.

Amy J. Ko and Katie Davis

Computing Mentorship ICER ’17, August 18-20, 2017, Tacoma, WA, USA

REFERENCES[1] Tammy D. Allen and Lillian T. Eby. 2011. �e Blackwell handbook of mentoring:

A multiple perspectives approach. John Wiley & Sons, USA.[2] Joshua D Angrist and Victor Lavy. 1999. Using Maimonides’ rule to estimate the

e�ect of class size on scholastic achievement. �e�arterly Journal of Economics114, 2 (1999), 533–575.

[3] Louise Archer and Hiromi Yamashita. 2003. ’Knowing their limits’? Identities, in-equalities and inner city school leavers’ post-16 aspirations. Journal of EducationPolicy 18, 1 (2003), 53–69.

[4] Brigid Barron, Caitlin Kennedy Martin, Lori Takeuchi, and Rachel Fithian. 2009.Parents as learning partners in the development of technological �uency. Inter-national Journal of Learning and Media 1, 2 (2009), 55–77.

[5] Margaret R Beam, Chuansheng Chen, and Ellen Greenberger. 2002. �e natureof adolescents’ relationships with their ”very important” non-parental adults.American Journal of Community Psychology 30, 2 (2002), 305–325.

[6] Lori Carter. 2006. Why students with an apparent aptitude for computer sciencedon’t choose to major in computer science. SIGCSE Bulletin 38, 1 (2006), 27–31.

[7] Gwo-Dong Chen, Liang-Yi Li, and Chin-Yea Wang. 2012. A Community of Prac-tice Approach to Learning Programming. Turkish Online Journal of EducationalTechnology-TOJET 11, 2 (2012), 15–26.

[8] Sapna Cheryan, Benjamin J Drury, and Marissa Vichayapai. 2013. Enduringin�uence of stereotypical computer science role models on women�s academicaspirations. Psychology of Women�arterly 37, 1 (2013), 72–79.

[9] Je�rey Cornelius-White. 2007. Learner-centered teacher-student relationshipsare e�ective: A meta-analysis. Review of educational research 77, 1 (2007), 113–143.

[10] A Craig. 1998. Peer mentoring female computing students–does it make adi�erence?. In Proceedings of the 3rd Australasian conference on Computer scienceeducation. ACM, USA, 41–47.

[11] Gloria Crisp and Irene Cruz. 2009. Mentoring college students: A critical reviewof the literature between 1990 and 2007. Research in higher education 50, 6 (2009),525–545.

[12] Phillip Dawson. 2014. Beyond a de�nition: Toward a framework for designingand specifying mentoring models. Educational Researcher 43, 3 (2014), 137–145.

[13] Betsy DiSalvo, Mark Guzdial, Amy Bruckman, and Tom McKlin. 2014. Saving facewhile geeking out: Video game testing as a justi�cation for learning computerscience. Journal of the Learning Sciences 23, 3 (2014), 272–315.

[14] Peggy Doerschuk. 2004. A research and mentoring program for undergraduatewomen in computer science. In Frontiers in Education. IEEE, USA, S2H–7–12.

[15] Benjamin J Drury, John Oliver Siy, and Sapna Cheryan. 2011. When do femalerole models bene�t women? �e importance of di�erentiating recruitment fromretention in STEM. Psychological Inquiry 22, 4 (2011), 265–269.

[16] Daryl D’Souza, Margaret Hamilton, James Harland, Peter Muir, Charles �e-vathayan, and Cecily Walker. 2008. Transforming learning of programming:a mentoring project. In Conference on Australasian Computing Education. Aus-tralian Computer Society, Inc., Australia, 75–84.

[17] Eric F Dubow, Paul Boxer, and L Rowell Huesmann. 2009. Long-term e�ects of par-ents� education on children�s educational and occupational success: Mediationby family interactions, child aggression, and teenage aspirations. Merrill-Palmerquarterly (Wayne State University. Press) 55, 3 (2009), 224.

[18] Carolyn M Evertson and Carol S Weinstein. 2013. Handbook of classroom man-agement: Research, practice, and contemporary issues. Routledge, England.

[19] Arline T Geronimus, John Bound, and Lisa J Neidert. 1996. On the validity of usingcensus geocode characteristics to proxy individual socioeconomic characteristics.J. Amer. Statist. Assoc. 91, 434 (1996), 529–537.

[20] Tony Greening. 1998. Computer science: through the eyes of potential students.In Australasian conference on Computer science education. ACM, New York, 145–154.

[21] Denise Gurer and Tracy Camp. 2002. An ACM-W literature review on womenin computing. ACM SIGCSE Bulletin 34, 2 (2002), 121–127.

[22] Mark Guzdial and Allison Ellio� Tew. 2006. Imagineering inauthentic legiti-mate peripheral participation: an instructional design approach for motivatingcomputing education. In ACM International Workshop on Computing EducationResearch. ACM, New York, 51–58.

[23] John C. Hall. 2003. Mentoring and Young People: A Literature Review. ERIC, USA.[24] Michael Hewner. 2013. Undergraduate conceptions of the �eld of computer

science. In ACM International Computing Education Research. ACM, New York,107–114.

[25] Michael Hewner and Mark Guzdial. 2008. A�itudes about computing in post-secondary graduates. In ACM International Workshop on Computing EducationResearch. ACM, New York, 71–78.

[26] Suzanne Hidi and K Ann Renninger. 2006. �e four-phase model of interestdevelopment. Educational psychologist 41, 2 (2006), 111–127.

[27] Mark A Holliday and David R Luginbuhl. 2004. Peer-centered service learning.In Frontiers in Education. IEEE, USA, F3E–1–6.

[28] Beth K Humberd and Elizabeth D Rouse. 2016. Seeing you in me and me in you:Personal identi�cation in the phases of mentoring relationships. Academy of

Management Review 41, 3 (2016), 435–455.[29] David Marshall Hunt and Carol Michael. 1983. Mentorship: A career training

and development tool. Academy of management Review 8, 3 (1983), 475–485.[30] �omas E. Keller. 2005. �e stages and development of mentoring relationships.

Handbook of youth mentoring 1 (2005), 82–99.[31] Harold R Kerbo. 1996. Social strati�cation and inequality: Class con�ict in historical

and comparative perspective. McGraw-Hill College, New York.[32] Rhodes Jean E. Klaw, Elena L. and Louise F. Fitzgerald. 2003. Natural mentors in

the lives of African American adolescent mothers: Tracking relationships overtime. Journal of Youth and Adolescence 32, 3 (2003), 223–232.

[34] Jean Lave and Etienne Wenger. 1991. Situated learning: Legitimate peripheralparticipation. Cambridge university press, Cambridge.

[35] Colleen M Lewis, Ruth E Anderson, and Ken Yasuhara. 2016. I Don’t CodeAll Day: Fi�ing in Computer Science When the Stereotypes Don’t Fit. In ACMConference on International Computing Education Research. ACM, New York,23–32.

[36] Jane Margolis, Rachel Estrella, Joanna Goode, Jennifer Jellison Holme, and KimNao. 2010. Stuck in the shallow end: Education, race, and computing. MIT Press,Cambridge, MA.

[37] Jane Margolis and Allan Fisher. 2003. Unlocking the clubhouse: Women in com-puting. MIT press, Cambridge, MA.

[38] J Mitchell Miller, JC Barnes, Holly Ventura Miller, and Layla McKinnon. 2013.Exploring the link between mentoring program structure & success rates: Resultsfrom a national survey. American Journal of Criminal Justice 38, 3 (2013), 439–456.

[39] Emma Norling. 1995. Encouraging networking through informal mentoring: alook at a newly-established mentor scheme. In Second Australasian Women inEngineering Forum. RMIT, Australia.

[40] Youn Joo Oh, Yueming Jia, Mhora Lorentson, and Frank LaBanca. 2013. De-velopment of the educational and career interest scale in science, technology,and mathematics for high school students. Journal of Science Education andTechnology 22, 5 (2013), 780–790.

[41] Becky Wai-Ling Packard and Paul F Conway. 2006. Methodological choice andits consequences for possible selves research. Identity 6, 3 (2006), 251–271.

[42] SS Pisimisi and MG Ioannides. 2005. Developing mentoring relationships tosupport the careers of women in electrical engineering and computer technolo-gies. An analysis on mentors’ competencies. European journal of engineeringeducation 30, 4 (2005), 477–486.

[43] K Ann Renninger and Suzanne Hidi. 2011. Revisiting the conceptualization,measurement, and generation of interest. Educational Psychologist 46, 3 (2011),168–184.

[44] Ashley Robinson, Manuel A Perez-�inones, and Glenda Scales. 2015. Under-standing the a�itudes of African American middle school girls toward computerscience. In Research in Equity and Sustained Participation in Engineering, Com-puting, and Technology (RESPECT), 2015. IEEE, New York, 1–8.

[45] Amber Se�le, Sarah Pieczynski, Liz Friedman, and Nathan Kizior. 2013. Evalu-ating a Prospective Student Mentoring Program. In International Conference onFrontiers in Education: Computer Science and Computer Engineering (FECS). IEEE,USA, 1.

[46] Selcuk R Sirin. 2005. Socioeconomic status and academic achievement: A meta-analytic review of research. Review of educational research 75, 3 (2005), 417–453.

[47] Claude M Steele and Joshua Aronson. 1995. Stereotype threat and the intellec-tual test performance of African Americans. Journal of personality and socialpsychology 69, 5 (1995), 797.

[48] Gloria Childress Townsend. 1996. Viewing video-taped role models improvesfemale a�itudes toward computer science. ACM SIGCSE Bulletin 28, 1 (1996),42–46.

[49] Gloria Childress Townsend. 2002. People who make a di�erence: mentors androle models. ACM SIGCSE Bulletin 34, 2 (2002), 57–61.

[50] Etienne Wenger, Richard Arnold McDermo�, and William Snyder. 2002. Cultivat-ing communities of practice: A guide to managing knowledge. Harvard BusinessPress, Cambridge, MA.

[51] Sarah Wood and Evan Mayo-Wilson. 2012. School-based mentoring for adoles-cents: A systematic review and meta-analysis. Research on Social Work Practice22, 3 (2012), 257–269.

[52] Sarita Yardi and Amy Bruckman. 2007. What is computing?: bridging the gapbetween teenagers’ perceptions and graduate students’ experiences. In ACMInternational Workshop on Computing Education Research. ACM, New York, 39–50.

[33] Amy J Ko. 2009. A�itudes and self-e�cacy in young adults’ computing autobiographies. In IEEE Symposium on Visual Languages and Human-Centric Computing. IEEE, New Jersey, 67–74.